The notion of interpreting cortical operations as probabilistic computation has been steadily gaining ground in neuroscience, and with the emergence of the PPC-based and sampling-based frameworks, now there exist clear theoretical alternatives of how such computation might happen in the brain. Nevertheless, a number of crucial issues necessary to make these frameworks biologically viable proposals of how the brain works remained unclear. I will focus on three such issues: providing a learning-based neurophysiological support of the probabilistic sampling framework, demonstrating a psychophysical hallmark of this framework, and finally, presenting behavioural evidence of optimal probabilistic computation at the lowest level of cortical visual processing.